tests/test_ast_extractor.py

"""M05 — Python AST Extractor. Red/green tests derived from the module spec.

Chosen public API (documented here, asserted below)::

    extract_python_file(
        source: str,
        relative_path: str,
        *,
        source_id: str = "",
    ) -> tuple[list[Node], list[Edge], list[RawImport]]

``source_id`` defaults to the normalised ``relative_path`` when empty. Extraction is
pure on text: ``SyntaxError`` is caught internally and yields ``([], [], [])``. A
``PythonExtractor(ast.NodeVisitor)`` with a scope stack does the work internally; a
``read_source(path)`` helper enforces the 1 MiB / encoding skip rules for callers.

Metadata stringification convention (``Node.metadata`` is ``dict[str, str]``):
- Booleans -> the lowercase strings ``"true"`` / ``"false"``
  (``is_method``, ``is_async``, ``is_property``).
- Lists -> compact ``json.dumps`` (``lambda_names``, ``pytest_markers``).
"""

from __future__ import annotations

import json
from pathlib import Path

from loosecanvas.contracts import Edge, Node, RawImport
from loosecanvas.extractors.ast_extractor import extract_python_file, read_source

_REL = "src/auth/providers.py"

SAMPLE = '''\
import os
from collections import OrderedDict
from . import sibling
from .pkg import thing as t
from m import *
from x import a, b


class Animal:
    """base"""

    def __init__(self, name):
        self.name = name

    @property
    def label(self):
        return self.name

    async def fetch(self):
        return 1

    class Meta:
        ordering = "name"


class Dog(Animal):
    def bark(self):
        adder = lambda value: value + 1  # noqa: E731
        return adder(1)
'''


def _ids(nodes: list[Node]) -> list[str]:
    return [n.id for n in nodes]


def _by_id(nodes: list[Node], node_id: str) -> Node:
    for n in nodes:
        if n.id == node_id:
            return n
    raise AssertionError(f"node not found: {node_id}\navailable: {_ids(nodes)}")


def _has_edge(edges: list[Edge], source: str, target: str, kind: str) -> bool:
    return any(
        e.source == source and e.target == target and e.kind == kind for e in edges
    )


def test_class_method_and_async_nodes() -> None:
    nodes, _edges, _imports = extract_python_file(SAMPLE, _REL)
    ids = set(_ids(nodes))

    assert f"{_REL}::class::Animal" in ids
    assert f"{_REL}::function::Animal.__init__" in ids
    assert f"{_REL}::function::Animal.label" in ids
    assert f"{_REL}::function::Animal.fetch" in ids

    init = _by_id(nodes, f"{_REL}::function::Animal.__init__")
    assert init.kind == "function"
    assert init.metadata["is_method"] == "true"
    assert init.metadata["is_async"] == "false"

    fetch = _by_id(nodes, f"{_REL}::function::Animal.fetch")
    assert fetch.metadata["is_async"] == "true"
    assert fetch.metadata["is_method"] == "true"


def test_nested_class_qualified_name_uses_dot() -> None:
    nodes, _edges, _imports = extract_python_file(SAMPLE, _REL)
    assert f"{_REL}::class::Animal.Meta" in set(_ids(nodes))


def test_property_metadata() -> None:
    nodes, _edges, _imports = extract_python_file(SAMPLE, _REL)
    label = _by_id(nodes, f"{_REL}::function::Animal.label")
    assert label.metadata["is_property"] == "true"


def test_inherits_edge() -> None:
    nodes, edges, _imports = extract_python_file(SAMPLE, _REL)
    dog = f"{_REL}::class::Dog"
    assert dog in set(_ids(nodes))
    assert any(
        e.source == dog and e.kind == "inherits" and "Animal" in e.target for e in edges
    )


def test_contains_edges_link_module_class_function() -> None:
    nodes, edges, _imports = extract_python_file(SAMPLE, _REL)
    module_id = f"{_REL}::module::providers"
    animal = f"{_REL}::class::Animal"
    init = f"{_REL}::function::Animal.__init__"
    assert module_id in set(_ids(nodes))
    assert _has_edge(edges, module_id, animal, "contains")
    assert _has_edge(edges, animal, init, "contains")


def test_lambda_names_recorded_on_parent() -> None:
    nodes, _edges, _imports = extract_python_file(SAMPLE, _REL)
    bark = _by_id(nodes, f"{_REL}::function::Dog.bark")
    assert json.loads(bark.metadata["lambda_names"]) == ["adder"]


def test_imports_collected_as_rawimport() -> None:
    _nodes, _edges, imports = extract_python_file(SAMPLE, _REL)

    def find(pred: object) -> RawImport:  # noqa: ANN401 - test helper
        return next(i for i in imports if pred(i))  # type: ignore[operator]

    os_imp = find(lambda i: i.module == "os")
    assert os_imp.names == [] and os_imp.level == 0

    from_x = find(lambda i: i.module == "x")
    assert from_x.names == ["a", "b"] and from_x.level == 0

    rel = find(lambda i: i.level == 1 and i.module == "")
    assert rel.names == ["sibling"]

    star = find(lambda i: i.names == ["*"])
    assert star.module == "m"

    assert all(i.current_file == _REL for i in imports)


def test_parametrize_marker_recorded() -> None:
    src = (
        "import pytest\n\n\n"
        '@pytest.mark.parametrize("x", [1, 2])\n'
        "def test_thing(x):\n"
        "    assert x\n"
    )
    nodes, _edges, _imports = extract_python_file(src, "tests/test_thing.py")
    node = _by_id(nodes, "tests/test_thing.py::function::test_thing")
    assert "parametrize" in json.loads(node.metadata["pytest_markers"])


def test_every_node_has_valid_evidence() -> None:
    nodes, _edges, _imports = extract_python_file(SAMPLE, _REL)
    assert nodes
    for n in nodes:
        assert len(n.evidence_refs) >= 1
        ev = n.evidence_refs[0]
        assert ev.snippet_hash
        assert ev.start_line <= ev.end_line


def test_node_ids_have_no_backslash_and_three_parts() -> None:
    nodes, _edges, _imports = extract_python_file(SAMPLE, "src\\auth\\providers.py")
    for n in nodes:
        assert "\\" not in n.id
        assert len(n.id.split("::")) == 3


def test_syntax_error_handled_gracefully() -> None:
    nodes, edges, imports = extract_python_file("def broken(:\n    pass\n", _REL)
    assert nodes == []
    assert edges == []
    assert imports == []


def test_determinism_no_id_collisions() -> None:
    first, _e1, _i1 = extract_python_file(SAMPLE, _REL)
    second, _e2, _i2 = extract_python_file(SAMPLE, _REL)
    ids = _ids(first)
    assert ids == _ids(second)
    assert len(ids) == len(set(ids))


def test_init_py_dual_nodes(tmp_path: Path) -> None:
    nodes, _edges, _imports = extract_python_file("x = 1\n", "pkg/sub/__init__.py")
    ids = set(_ids(nodes))
    # module node uses the package (parent dir) qualified name, never "__init__".
    assert "pkg/sub/__init__.py::module::sub" in ids
    assert not any(i.endswith("::module::__init__") for i in ids)


def test_read_source_skips_large(tmp_path: Path) -> None:
    big = tmp_path / "big.py"
    big.write_bytes(b"# " + b"x" * 1_100_000)
    assert read_source(big) is None


def test_read_source_latin1_fallback(tmp_path: Path) -> None:
    p = tmp_path / "lat.py"
    p.write_bytes(b"# calf\xe9\n")  # invalid utf-8, valid latin-1
    text = read_source(p)
    assert text is not None and "calf" in text